karthink

If you're adding files non-interactively, you can just manipulate `gptel-context` directly? It's a user option now, not an internal variable, so the interface should be stable. ``` emacs-lisp (cl-pushnew (expand-file-name...

> So now I noticed something strange. If I edit the context every time I press C-c > C-c in context-mode the context reverses order. > I thought the context...

> nfedyashev created an issue (karthink/gptel#1123) > > I always review/update/partially copy LLM respones in the *scratch* buffer before pasting them to files. > > it is a bit inconvenient...

> next time I run gptel - output defaults to "insert response at point" instead of scratch buffer. Not sure what I did wrong Well that sounds like a bug...

> I wish simple default settings could be manageable via variables instead of only via shortcuts. There are too many gptel settings (easily hundreds) to have variables for all of...

@aagit I think you've sent me emails with a couple of patches about a similar issue that I haven't got to yet. Do you want me to ignore the emails...

@aagit Two notes: 1. The right way to do this would be to replace `gptel--rewrite-directive` with `(gptel--parse-directive gptel--rewrite-directive 'raw)` instead of checking if it's a function here. (The directive can...

After some more thinking, it makes sense for the system message to always be parsed (eval'd if it's a function) in the original buffer -- if it's a string or...

I've had more time to ponder this behavior now, and I think the correct thing to do is to parse the directive at the time of request, in the context...

It's just a little tricky to do correctly since the directive can take many forms -- a string, list, or a function that can return a string or a list....